. 2022 Jun 16;2022:9414567. doi: 10.1155/2022/9414567

Table 3.

Sizes of outputs and convolutional kernels for ResNet versions.

Layer name	Output size	34 layers	50 layers	101 layers
conv 1	112 × 112	7 × 7, 64, stride 2
conv 2.x	56 × 56	3 × 3 max pool, stride 2
conv 2.x	56 × 56	$[\begin{matrix} 3 x 3,64 \\ 3 x 3,64 \end{matrix}] x 3$	$[\begin{matrix} 1 x 1,64 \\ 3 x 3,64 \\ 1 x 1,256 \end{matrix}] x 3$	$[\begin{matrix} 1 x 1,64 \\ 3 x 3,64 \\ 1 x 1,256 \end{matrix}] x 3$
conv 3.x	28 × 28	$[\begin{matrix} 3 x 3,128 \\ 3 x 3,128 \end{matrix}] x 4$	$[\begin{matrix} 1 x 1,128 \\ 3 x 3,128 \\ 1 x 1,512 \end{matrix}] x 4$	$[\begin{matrix} 1 x 1,128 \\ 3 x 3,128 \\ 1 x 1,512 \end{matrix}] x 4$
conv 4.x	14 × 14	$[\begin{matrix} 3 x 3,256 \\ 3 x 3,256 \end{matrix}] x 6$	$[\begin{matrix} 1 x 1,256 \\ 3 x 3,256 \\ 1 x 1,1024 \end{matrix}] x 6$	$[\begin{matrix} 1 x 1,256 \\ 3 x 3,256 \\ 1 x 1,1024 \end{matrix}] x 23$
conv 4.x	7 × 7	$[\begin{matrix} 3 x 3,512 \\ 3 x 3,512 \end{matrix}] x 3$	$[\begin{matrix} 1 x 1,512 \\ 3 x 3,512 \\ 1 x 1,2048 \end{matrix}] x 3$	$[\begin{matrix} 1 x 1,512 \\ 3 x 3,512 \\ 1 x 1,2048 \end{matrix}] x 3$
	1 × 1	Average pool, 1000-d fc, softmax
FLOPs		3.6 × 10⁹	3.8 × 10⁹	7.6 × 10⁹